| license: other | |
| license_name: circlestone-labs-non-commercial | |
| tags: | |
| - text-to-image | |
| - diffusers | |
| - cosmos | |
| library_name: diffusers | |
| pipeline_tag: text-to-image | |
| # Anima Preview 3 (SD.Next Diffusers Conversion) | |
| Diffusers-format conversion of [Anima Preview 3](https://huggingface.co/circlestone-labs/Anima) for use with SD.Next. | |
| **Original model:** [circlestone-labs/Anima](https://huggingface.co/circlestone-labs/Anima) | |
| ## Changes from Preview 2 | |
| - Extended training at 1024 resolution | |
| - Expanded dataset coverage for less common artists (50-100 post count) | |
| ## Architecture | |
| - **Transformer:** CosmosTransformer3DModel (2B params, 28 layers) | |
| - **Text Encoder:** Qwen3-0.6B (replacing Cosmos T5-11B) | |
| - **LLM Adapter:** Custom cross-attention adapter bridging Qwen3 to the transformer | |
| - **VAE:** AutoencoderKLWan | |